סמינר של בית הספר להנדסה מכאנית

יום ב' 12 במרץ 2012, שעה 15:00,

חדר 206, בניין וולפסון להנדסה מכאנית

Evolutionary Search of Near-optimal Concepts

based on Relaxed-Pareto-Optimality

Elad Denenberg

School of Mechanical Engineering, Faculty of Engineering

Tel Aviv University

This study presents a method of utilizing Evolutionary Algorithms (EAs) to support concept selection under conflicting objectives. It concerns a recent idea of concept-based relaxed-Pareto-optimality, which is employed to develop a "soft" set-based concept optimization and its related evolutionary search approach.  Relaxed-optimality aims at supporting designers and planners, which are engaged in concept selection. It helps examining concepts and comparing them in an exploratory way rather than under strict optimization.

The proposed search method treats sets of particular solutions simultaneously, where each set represents a concept. It uses the search drive to find the concept-based Pareto optimal set. However, it allows solutions, with performances close to those of the concept-based Pareto-optimal set, to survive the evolutionary search process. The numerical approach involves a modification of two EAs which are traditionally used to solve multi-objective problems: NSGA-II and ε-MOEA. The modification is twofold: modifying the algorithms to accommodate concept-based search, and using the new relaxed optimality definition to meet the needs of solving the described problem. The suggested algorithm is demonstrated using both academic test functions and a robot path planning problem.

This presentation is of an MSc thesis supervised by Dr. Amiram Moshaiov